This week’s stream of analyst notes reflected a market in the process of repricing winners and recalibrating risks as AI-related demand reshapes parts of the technology supply chain. Brokers differed sharply on where value now sits: some elevated long-term production and design leaders, while others warned that rapid share-price gains and shifting seasonal patterns may have outpaced fundamentals.
Apple valuation flagged as “stretched” by KeyBanc
KeyBanc Capital Markets told clients it sees mounting questions around Apple’s valuation after its proprietary spending data showed signs of weaker-than-usual U.S. hardware demand. Analyst Brandon Nispel kept a Sector Weight rating on the stock and pointed to field-level metrics indicating indexed spending fell 16% month-over-month in April, deeper than the three-year average decline of 12%.
Nispel’s note highlighted a sharp deceleration in year-on-year growth for indexed spending, which slid to negative 6% from positive 10% the prior month. The bank said this pattern points to an easing of the robust hardware momentum that underpinned recent optimism, and it underscores the growing importance of international markets - particularly China and Asia-Pacific - for future growth assumptions, areas where KeyBanc acknowledged it has less direct visibility.
Despite expressing concern, KeyBanc’s near-term modeling remains above consensus for hardware revenue in Apple’s fiscal third quarter: the firm projects hardware revenue growth of 18.2% year-over-year versus a consensus of 15.4%. Within that figure, KeyBanc models iPhone revenue rising 24.6% compared with a consensus estimate of 19.4%.
Even with these above-consensus near-term estimates, the bank said it is having difficulty reconciling the forecasts with Apple’s current valuation. KeyBanc noted that Apple is trading at roughly 23 times fiscal 2027 EV/EBITDA and 32 times earnings, and that the stock appears generally high relative to its historical discount to the Nasdaq and peers among large-cap technology names. Additional cautionary points included Apple’s business cyclicality, U.S. carrier signals suggesting fewer device subsidies, and commentary from Verizon indicating upgrade rates are moderating.
BofA reiterates Buy on TSMC, argues competitive concerns are overstated
Bank of America maintained a Buy rating and a price target of NT$2,560 on Taiwan Semiconductor Manufacturing Company after the chipmaker’s technology symposium in Taiwan. BofA framed recent market worries as overblown, arguing that TSMC’s scale and lead in manufacturing technology are expanding, not contracting.
On advanced nodes, the bank detailed TSMC’s capacity targets and anticipated throughput. For the 3nm and 5nm families, TSMC is targeting a compound annual capacity growth rate of 25% from 2022 to 2027, with 3nm capacity projected to reach 190,000 wafers per month by the fourth quarter of 2026 and to expand to 230,000 wafers per month by 2027.
BofA contrasted TSMC’s scale with estimates for competing processes - Samsung’s SF3 and Intel’s 18A - which it places at roughly 20,000 to 25,000 wafers per month at early, low yields and primarily earmarked for internal use. The bank said that disparity materially limits the risk that major customers will switch away from TSMC for critical chipsets.
Looking forward to TSMC’s next-generation N2 node, BofA noted a targeted 70% compound annual capacity growth between 2026 and 2028. That program includes simultaneous deployment of five fabrication facilities and an objective of cutting technology transfer times by 20%. BofA also highlighted that N2 achieved its targeted defect density two quarters ahead of the timetable set during the 3nm ramp.
On packaging, TSMC is planning aggressive capacity expansion for CoWoS and SoIC, with compound annual growth rates of approximately 80% and 90% through 2027. The bank pointed out that CoWoS yields at TSMC are already above 98%, while a competing packaging approach referenced in the note - Intel’s EMIB-T - remains at pilot yields of 80% to 85%. BofA said the yield gap creates execution risk for the competitor if it cannot reach roughly 95% mass-production yield by mid-2027.
KB Securities lifts SK Hynix and Samsung targets as memory shortage risk intensifies
KB Securities raised its price targets for SK Hynix and Samsung Electronics on the view that AI infrastructure investment is producing a structural supply shortfall for memory products. For SK Hynix the broker increased its target to 3,000,000 won from 2,800,000 won while keeping a Buy rating; Samsung was upgraded to Buy with a fresh target of 450,000 won.
Analyst Jeff Kim framed hyperscaler spending as a strategic race for AI infrastructure leadership rather than a simple capex contest. He argued that hyperscale cloud operators will continue investing heavily until sufficient memory capacity and complementary data center infrastructure are in place. KB’s research projects that AI-related token usage at major cloud platforms will rise about three times over the next six months and seven times year-on-year, creating acute near-term pressure on memory supply.
The note quantified current demand patterns, observing that AI data center operators already consume roughly 70% of total memory shipments. KB said that demand is set to broaden as agentic AI expands into on-device and physical AI use cases.
Importantly for supply-side dynamics, KB warned that new memory production lines are not slated for mass production until after 2027, a timeline that the broker said implies the industry could be entering a de facto zero-supply era. In its pricing forecasts, KB projects SK Hynix’s 2026 DRAM and NAND average selling prices to surge 194% and 244% year-on-year respectively, and it models similarly steep gains for Samsung of 297% and 256%.
On profitability, KB sees an exceptionally profitable 2026 for SK Hynix: operating profit of 277 trillion won at an operating margin of 78.1%, which the broker said would rank the company fourth globally by that metric. It also forecast second-quarter operating profit jumping more than eight times year-on-year to 70 trillion won.
KB highlighted strategic shifts as well, noting that SK Hynix’s move toward longer-term supply contracts covering 2028 through 2030 increasingly resembles a foundry-like model: advance orders followed by production. The broker said such arrangements can reduce earnings volatility and improve visibility, which in turn may drive multiple re-rating.
Backing these company-level projections is a pronounced acceleration in hyperscaler capital expenditure, KB estimated. The broker expects combined capex from four large U.S. cloud operators - including the two named in the note - to reach $725 billion in 2026, up 77% year-on-year, and to surpass $1 trillion in 2027. KB summarized the strategic dynamic bluntly: for the largest technology platforms, AI capex is becoming a survival barrier rather than a discretionary cost.
Daiwa trims AMD recommendation after strong share run
Daiwa Capital Markets moved AMD from Buy to Outperform even as it raised the stock’s price target to $500 from $250, following a strong first-quarter report. The analyst’s action reflects a distinction between the company’s operational strength and the stock’s recent market performance.
Analyst Louis Miscioscia described AMD’s first-quarter 2026 results as "very good," pointing to revenue of $10.3 billion, a 38% year-on-year increase that beat the Street estimate of $9.9 billion by $361 million. Gross margins came in at 55.4%, slightly above guidance. AMD’s second-quarter revenue guidance of $11.2 billion exceeded consensus by $682 million and implied year-on-year growth of 46%.
Management concurrently pushed out a more bullish long-term view, doubling its 2030 estimate for the x86 total addressable market to more than $120 billion and lifting its long-range compound annual growth rate projection to 35% from 18%. The company also set a target of exceeding $20 in earnings per share within a three-to-five-year horizon.
Despite the optimistic operational outlook and the higher estimates Daiwa applied to revenue and earnings - $49 billion and $7.25 in 2026, and $72.7 billion and $12.85 in 2027 - the brokerage sought to temper near-term investor expectations. Miscioscia cited the stock’s appreciation of almost 150% over the previous 60 days and suggested the pace of gains could moderate in the near term, which was the principal reason for lowering the recommendation.
HSBC upgrades Cisco, arguing AI demand alters growth profile
HSBC raised its rating on Cisco to Buy and increased its price target to $137 from $77, asserting that rising AI infrastructure orders have changed the networking company’s growth trajectory. The bank pointed to a marked structural shift in Cisco’s recent results, where AI infrastructure orders reached $1.9 billion in the quarter compared with $600 million a year earlier, lifting the year-to-date total to $5.3 billion.
Management revised its fiscal targets upward: Cisco increased its fiscal 2026 AI infrastructure orders target to about $9 billion from $5 billion and raised full-year AI revenue guidance to roughly $4 billion from $3 billion. For fiscal 2027, management guided for AI revenue of at least $6 billion - implying about 50% year-on-year growth and total company revenue growth near 9%.
HSBC credited Cisco’s performance to product and design wins, specifically calling out the company’s Silicon One chips and Acacia optical products supported by multiple hyperscaler design wins. Cisco’s total revenue grew 12% year-on-year to $15.84 billion, and adjusted earnings per share of $1.06 beat estimates by 2.5%.
In light of those developments, HSBC raised its non-GAAP EPS compound annual growth rate for fiscal 2026 through 2029 to 13.6% from 9.8% and applied a higher 29 times price-to-earnings multiple to reflect what it described as a transition from a low-growth networking company to a structural AI infrastructure story. The bank identified three drivers underpinning the upgrade: hyperscaler AI buildouts, enterprise AI networking upgrades, and broader campus modernization driven by rising traffic and security requirements.
Implications across sectors
The analyst actions this week touch multiple parts of the market. Semiconductor fabrication and packaging capacity planning, memory supply economics and hyperscaler capex cycles directly influence chipmakers and equipment suppliers. Networking vendors and infrastructure suppliers stand to gain where AI order momentum persists. Conversely, hardware OEMs that rely on cyclical consumer upgrade patterns face renewed scrutiny when field-level spending data softens.
These notes collectively underline how AI investment is shifting demand profiles across chips, memory, packaging and networking - creating winners for scale and execution while elevating valuation and supply-timing risk where expectations have raced ahead of visible supply additions.
Conclusion
Broker research this week reflects a market sorting winners and risks as AI deployment accelerates. Banks rewarded scale and execution - particularly at TSMC, and in networking at Cisco - while also lifting price targets for memory suppliers on a forecasted supply squeeze. At the same time, analysts signaled caution where stretched valuations and recent rapid share-price moves may have outrun the near-term fundamentals. The net result is a mosaic of upgrades and prudent downgrades that emphasize both the transformative demand from AI and the uneven ability of suppliers and customers to match that demand on favorable economics.
Investors and market participants should therefore reckon with both structural demand shifts tied to AI buildouts and the timing of supply responses, while keeping an eye on valuation compression where share prices have advanced rapidly.